The PMOD MACHINE LEARNING workflow supports the learning as well as the prediction part. The users can prepare sets of images with their corresponding segmentation. The learning sets can be submitted to TensorFlow for training a neural network of suitable structure. The resulting network can finally be applied to segment input images with characteristics similar to the ones in the learning set.
Please refer to the PAI Documentation for details about the creation of a learning set, the training of a network, the prediction-based image segmentation, as well for all information related to the infrastructure requirements.